Updated with data available as of 2020-05-20
Aims
The USC Predict COVID project is using an epidemic model to estimate the impact of COVID-19 in Los Angeles County
- We are addressing the key questions of:
- When will the peak of the epidemic occur and how will it impact health care capacity?
- What happens to the dynamics of the epidemic when social distancing ends?
- How will the epidemic affect different at-risk groups?
- Critical healthcare variables predicted by the model are the counts of the numbers of individuals over time, including the peak occurrence, for the following:
- The total number of infected cases including both the number detected and observed with testing and the undetected/untested cases
- The total number of individuals hospitalized (including those in the ICU)
- The number of patients in the ICU
- The number of patients on ventilators
- The number of deaths
- We estimate a number of key epidemic parameters, including:
- \(R0\), the reproductive number or average number of new infections generated by an infected person in a completely susceptible population
- \(r\), the proportion of illnesses that are detected and reported out of all illnesses. In this document we will consider that a prior for \(r\) can be directly informed by seroprevalence studies.
- \(Frac_{R0}\), the factor reduction in the initial R0 due to social distancing
- \(Pr(Hospital | Illness)\), the probability of hospitalization given illness
- \(Pr(Ventilation | ICU)\), the probability of ICU care necessary given hospitalization
- \(p_v\), the probability of ventilation given ICU care
- \(Pr(Death | ICU)\), the probability of death given ICU care
- We provide projections of illness severity trajectories in L.A. County as a whole and for race/ethnicity groups, based on prevalence of known COVID-19 risk factors.
- We provide predictions for the impact on counts and corresponding time periods under various social distancing scenarios in which restrictions are eased.
(1) Model projections
(1.1) Projections Against Data
Demonstrating model fit against COVID-19 data for Los Angeles, for the following variables/compartments:
- New = new daily incidence
- Current = current census in compartment
- Cumulative = running total over time
COVID-19 data is shown as black dots in the figures below.
Short time horizon

Longer time horizon

{-}
(1.2) Detailed Projections by Key Compartments
Projections for key compartments:
- Current census in Hosptial
- Current census in ICU
- Current census in Ventilation
- Cumulative deaths
- Current Illnesses: Observed
- Current Illnesses: Total (observed + unobserved)
Current Observed Illnesses
Current Total Illnesses (Observed + Unobserved)
{-}
(1.3) Projections for LA County and by SPA Given Risk Factors
Risk Profiles, Risk Factors, and Risk Groups
- The following table presents the model-estimated probabilities \(Pr(Hospital | Illness,Profile_i)\), \(Pr(ICU | Hospital,Profile_i)\), and \(Pr(Death | ICU,Profile_i)\) for each risk group (or combination of risk factors), as well as the prevalence of these risk groups/factors in the general L.A. County Population \(Pr(Profile_i)\).
Projections by Risk Factors
- These figures show the estimated proportion of each risk group that will make up the resulting cohorts of COVID patients admitted to hospital, admitted to ICU, or that die within the L.A. County/SPA population, based on the population prevalence of the risk group in L.A. County/SPAs.
- The analyses are presented for each risk group, as well as stratified to the individual risk factors (age, comorbidities, obesity status, smoking status).
By SPA, stage of disease

By stage of disease, SPA

Comparing risk factors

Age by stage of disease

Comorbidity by stage of disease

BMI by stage of disease

Smoking by stage of disease

Projections across risk factors: Fraction of all LA County Cases by risk factor
By risk group

By age

By comorbidity

By obesity status

By smoking status

(1.4) Projections by Race / Ethnicity
Projections by race/ethnicity
- These figures show the estimated proportion of each risk group that will make up the resulting cohorts of COVID patients admitted to hospital, admitted to ICU, or that die within each race/ethnicity population, based on the population prevalence of the risk group in race/ethnicity groups.
- The analyses are presented for each risk group, as well as stratified to the individual risk factors (age, comorbidities, obesity status, smoking status).
Risk groups by stage of disease, race

Risk groups by race, stage of disease

Age by stage of disease, race

Comorbidity by stage of disease, race

BMI by stage of disease, race

Smoking by stage of disease, race

Projections across race/ethnicity: Fraction of all LA County Cases by Race
- Due to disparities in the prevalence of health conditions, health behaviors, and age – the risk factors for COVID-19 – the relative risk of death is not distributed equally across each race/ethnicity.
- These figures demostrate the projections of the relative share of hospitalizations, ICU admissions, and deaths by race/ethnicity from our model, which accounts for biological COVID-19 risk factors. Values are compared with the baseline distribution of the relative share of the L.A. County population in each race/ethnicity group.
Deaths by race: Model projections vs. observed LA data
- This figure compares the projections of the relative share of deaths by race/ethnicity from our model with the observed relative share of deaths by race/ethnicity in L.A. County. The population prevalence of each race/ethnicity is also provided as a baseline (left column).
- A key insight from this figure is that as the epidemic evolves, the trend in death rate by race/ethnicity appears to be converging to the biological expected risk as calculated by our model.
- Only time will tell if this stabilies or if the death rate by race/ethnicity grows disproportionately to that expected by biological factors alone as the epidemic progresses.

Projected illness severity by race

Projected illness trajectories by race and risk group

(2) Social distancing scenarios
Summary
- Predictive epidemic modeling can help to evaluate the public health strategies for limiting the spread of COVID-19. LA County is succeeding in mitigating the epidemic curve due to strong adherence to social distancing (model-estimated reduction in contact rate of >70%).
- Our model suggests that social distancing may need to be continued for a longer period of time to retain these mitigation effects.
- Our model predicts that higher risk groups – characterized not only by advanced age but also combinations of existing health conditions – will make up the majority of those afflicted with severe disease. We analyze how these risk factors project onto race/ethnicity groups and project expected illness patterns based on population size and prevalence of these risk factors.
- Model estimates can and will change as new studies and data become available.
Acknowledgements
- We would like to acknowledge Claire Jacquillat and Adam Taylor for their contributions related to data wrangling, cleaning, interpretation, and visualization.
(2) Social distancing scenarios
SCENARIOS EVALUATED: Increase contact rate on May 16th, 2020 by variable amounts
Scenarios evaluated
7 are illustrated here. Each scenario corresponds to a different amount of contact rate, and can be understood in terms of:
Maintain current level of social distancing
Equal to:
% of original R0: 35.8
Equal to:
% of original R0: 48.7
Equal to:
% of original R0: 61.5
Equal to:
% of original R0: 74.3
Equal to:
% of original R0: 87.2
Equal to:
% of original R0: 100
Equal to: